I just tried testing the system one model because it's trending these days.
Roughly, is it correct that today is Tuesday in Korea and tomorrow is Wednesday? That kind of thing.

It's a system one model from Sola, 0.05 dollars per 1 million yes is about 84%.

It's a kev -4b model, 0.042 dollars per 1 million, the same price as jev, yes is 76.3%.

Jev, the representative model of system one, is 0.042 dollars per 1 million, yes is 96%.

Span-01 lite model, it's a lite model and like the free model, it was released for free, no 97%.

Span-01 model, 0.02 dollars per 1 million, similar results to the lite model, no 92.5%.
My settings might be wrong, or the model might not understand the problem I'm giving it.
It's not accurate. I don't want to say this is good or bad based on just one problem.
However, I tried using jev for 10 dollars, and so far among the publicly released models, jev seems to be the best model.
I believe that many more accurate and cheaper models will come out in the future.